The opening sentence of the homepage copy of two of London's newest luxury hotels — properties owned by different groups, designed by different agencies, opened five months apart — is, word for word, the same sentence. Neither marketing director chose it. Neither of them noticed until it was pointed out.
This is what AI failure in luxury actually looks like. Not the campaign disaster. Not the public misstep. The brand voice flattens by a few degrees. The content calendar fills out but the standout pieces disappear. The team produces more. It produces less of what mattered. None of it shows up in a report. All of it shows up in the brand, eventually. Not failure but erosion — invisible until it is advanced, cumulative rather than episodic, and by the time it is visible the damage is structural rather than correctable. A luxury brand can survive a campaign that misses. It cannot easily recover from two years of work that was almost, but not quite, its own.
i. What is already happening, quietly
In luxury hospitality, property and travel, there is no single backlash moment to point to. The work in question is not the campaign, and the work itself may not look like anything is wrong. It is appearing in the editorial newsletter, the brochure copy for a new residence, the longform blog content that drives organic search, the press materials going out under the brand's name, the captions for a seasonal campaign. Work that still requires brand judgement, but no longer receives it — not because the output is poor, but because no one in the brand has decided that the output is what the brand should say.
What is being seen, inside the function, is that AI has already arrived there. Not as policy. Not as a sanctioned tool with clear boundaries. As something the marketing manager reached for at four o'clock on a Friday because she had three other deliverables that week. As something the agency used on the third draft because the deadline moved. As something the junior team member is now better at using than anyone senior enough to govern it.
None of this is necessarily producing bad work. It is producing work that no one inside the brand has authored. And it is happening at scale — recent research found 49% of employees admit using AI tools their employer has not approved, and 60% agree the risk is worth it if it helps meet deadlines. Very few luxury organisations can accurately say where AI is in their output today, or who in the business is making the decisions about what it says.
ii. The visible failures are the tip
The pattern is easier to see in fashion, where AI has already produced public failures. Valentino faced widespread criticism in December 2025 after posting an AI-generated video on Instagram promoting its DeVain handbag, with users calling the imagery cheap, lazy and disturbing despite the brand clearly labelling the post as AI-generated. Two months later, Gucci faced similar backlash for releasing AI-generated promotional images ahead of its Milan Primavera show, with critics arguing the move into synthetic imagery contradicted the artisanal heritage the high prices are built on. Prada's Spring/Summer 2026 campaign drew the same response.
The pattern is consistent. Consumers see AI in luxury and read it as a downgrade. Even full transparency about AI use is not enough to recover the perception. A 2025 study in the Journal of Advertising Research, titled "When AI Doesn't Sell Prada", found that consumers exposed to AI-generated luxury advertising perceive less human effort behind the work, which undermines the authenticity that luxury pricing depends on.
What this hides is the more dangerous version of the same problem. Public AI failures get noticed because they happen in public. The quieter version is happening inside marketing functions every day — without surfacing as a campaign, without drawing comment, and without anyone at the top of the business knowing it has happened.
iii. What AI can do, and what it cannot
There is a serious argument against everything written above, and it deserves to be taken seriously.
The argument runs like this. AI tools are no longer the blunt instruments they were two years ago. A model with persistent memory, fine-tuned on a brand's archive, trained on its tone of voice documents, fed every brand book and campaign and press piece the brand has ever produced, will produce copy that passes for the brand. It will pass more reliably than a tired junior copywriter at the end of a long week. The model does not have bad days. It does not drift from the brief. It applies the voice with more consistency across more touchpoints than any human team could. Several major hospitality groups are already doing this, and the work is getting better, not worse. The premise that AI cannot capture brand voice is, on the technical question, no longer true.
This argument is correct, as far as it goes. The problem is that voice is not as far as it goes.
Brand voice is the surface. Underneath brand voice sits brand judgement — the thousand small decisions about what not to say, what not to publish, what not to lean into this season, what to hold back, what to leave unwritten, what to refuse. Brand judgement is the work of an editor, not a writer. Most luxury brands no longer employ either.
It is the work of leaving a flagship suite out of the website entirely, because it is held back for guests the brand chooses to extend it to. It is the work of refusing a partnership with a magazine every other property in the category has appeared in, because the magazine's audience is one step too wide.
A model can be trained on what a brand says. It cannot be trained on what a brand has chosen not to say, because those decisions do not appear in the archive. They are the absences in the archive — the campaigns refused, the words rejected, the opportunities turned down. They exist only in the heads of the people who made them, and increasingly, those people are no longer in the room when the work is being made.
This is the distinction the rest of the industry is not yet making. Voice is a problem AI has largely solved. Judgement is a problem AI cannot solve, because the data it would need to solve it does not exist in any retrievable form.
iv. The right question
The question luxury marketing functions are being sold on is the wrong one. Where can AI help us? is the question every AI agency is happy to answer, because it is the question every AI agency has been hired to answer.
The harder question — the one nobody is asking yet — is the one that follows from those two identical sentences. How long has the brand been speaking in a voice that is almost, but not quite, its own? And who in the organisation is left who would know the difference?